15N-{Ile, Leu, Val} into a Perdeuterated, 15N

Zwahlen, Kevin H. Gardner, Siddhartha P. Sarma, David A. Horita, R. Andrew Byrd, and Lewis E. Kay .... Scott A McCallum , T.Kevin Hitchens , Gordo...
69 downloads 0 Views 79KB Size
+

6800

+

J. Am. Chem. Soc. 1996, 118, 6800-6801

Incorporation of 1H/13C/15N-{Ile, Leu, Val} into a Perdeuterated, 15N-Labeled Protein: Potential in Structure Determination of Large Proteins by NMR William J. Metzler, Michael Wittekind, Valentina Goldfarb, Luciano Mueller, and Bennett T. Farmer II* Pharmaceutical Research Institute Bristol-Myers Squibb, P.O. Box 4000 Princeton, New Jersey 08543-4000 ReceiVed February 15, 1996 Proton-proton NOEs are the primary data for protein structure determination by NMR; it is therefore imperative that protons involved in these NOEs be accurately assigned. To this end, uniform 13C/15N isotopic labeling has become the mainstay for solution-state structural studies of both small and large proteins.1 For small proteins, such labeling often leads both to a faster2 and to a more precise determination of structure.3 For proteins in the range of 100-230 residues, uniform 13C/15N labeling is essential to NMR structural studies. With further increases in protein size, spectral dispersion yields to sensitivity as the limiting experimental factor. Substantial gains in sensitivity have been recently reported for many tripleresonance scalar correlation experiments on proteins deuterated at non-exchangeable proton sites.4,5 Although perdeuteration also increases sensitivity in amide-detected NOE experiments on large proteins,6-8 in solution-state structure determination, perdeuteration severely limits both the maximum number and the nature of possible proton-proton NOEs. In NOE experiments, random fractional deuteration (∼60%) of otherwise unlabeled proteins increases both general proton resolution and sensitivity.6,9 Similarly, random fractional deuteration limits spin diffusion by reducing the number of available magnetization-transfer pathways.6,9 This reduction permits a more efficacious use of longer NOE mixing times. With uniformly 13C-labeled proteins, however, random fractional deuteration will lead to a distribution in the total deuterium isotope effect4,10 on each 13C chemical shift. This distribution can severely broaden resonances along the 13C dimension. In addition to the random fractional deuteration of unlabeled proteins, techniques have also been described for type-specific protonation within an otherwise deuterated protein.11-13 In * To whom correspondence should be addressed. (1) Wittekind, M.; Mapelli, C.; Farmer, B. T., II; Suen, K. L.; Goldfarb, V.; Tsao, J.; Lavoie, T.; Barbacid, M.; Meyers, C. A.; Mueller, L. Biochemistry 1994, 33, 13531-13539. Metzler, W. J.; Farmer, B. T., II; Constantine, K. L.; Friedrichs, M. S.; Lavoie, T.; Mueller, L. Protein Sci. 1995, 4, 450-459. Tanaka, T.; Ames, J. B.; Harvey, T. S.; Stryer, L.; Ikura, M. Nature 1995, 376, 444-447. (2) Vis, H.; Boelens, R.; Mariani, M.; Stroop, R.; Vorgias, C. E.; Wilson, K. S.; Kaptein, R. Biochemistry 1994, 33, 14858-14870. (3) Clore, G. M.; Gronenborn, A. M. Science 1991, 252 (5011), 13909. (4) Grzesiek, S.; Anglister, J.; Ren, H.; Bax, A. J. Am. Chem. Soc. 1993, 115, 4369-4370. (5) Yamazaki, T.; Lee, W.; Arrowsmith, C. H.; Muhandiram, D. R.; Kay, L. E. J. Am. Chem. Soc. 1994, 116, 11655-11666. Venters, R. A.; Huang, C.-C.; Farmer, B. T., II; Trolard, R.; Spicer, L. D.; Fierke, C. A. J. Biomol. NMR 1995, 5, 339-344. (6) LeMaster, D. M.; Richards, F. M. Biochemistry 1988, 27, 142-50. Torchia, D. A.; Sparks, S. W.; Bax, A. J. Am. Chem. Soc. 1988, 110, 23202321. (7) Venters, R. A.; Metzler, W. J.; Spicer, L. D.; Mueller, L.; Farmer, B. T., II J. Am. Chem. Soc. 1995, 117, 9592-9593. (8) Grzesiek, S.; Wingfield, P.; Stahl, S.; Kaufman, J. D.; Bax, A. J. Am. Chem. Soc. 1995, 117, 9594-9595. (9) Torchia, D. A.; Sparks, S. W.; Bax, A. J. Am. Chem. Soc. 1988, 110, 2320-2321. (10) Hansen, P. E. Prog. NMR Spectrosc. 1988, 20, 207-255. (11) Crepsi, H. L.; Rosenberg, R. M.; Katz, J. J. Science 1968, 161, 795796. (12) Markley, J. L.; Potter, I.; Jardetzky, O. Science 1968, 161, 12491251.

contrast to random fractional deuteration, type-specific protonation preferentially increases local rather than global proton density; type-specific protonation therefore retains more of the benefits of perdeuteration. To circumvent the problems inherent in both the perdeuteration and the random fractional deuteration of uniformly 13C/ 15N-labeled proteins, the original technique for type-specific protonation has been modified to provide the following: 13C/ 15N labeling of the protonated specific residue types; uniform 15N labeling; and more complete deuteration of non-exchangeable sites on all other residue types. Furthermore, the modified expression medium does not require all amino-acid residue types; only the 1H/13C/15N-labeled specific residue types need be supplied. In this communication, we describe the methodology behind this novel approach to type-specific protonation; evaluate in a small protein the specificity and extent of protonation for one set of residue types ({Ile, Leu, Val} ≡ ILV); and demonstrate the potential benefits of such type-specific ILV protonation to the NMR structure determination of a large protein. Previous work has shown that a coarse global fold can be determined for a perdeuterated large protein (HCA II) using solely 1HN-1HN NOEs with a 6 Å cutoff distance.7 With this in mind, ILV residues have been chosen as the 1H/13C-labeled specific residues for the following three reasons: (1) favorable methyl-group T2 relaxation even with 13C labeling;14,15 (2) strong NOE interactions in the protein core; and (3) minimal scrambling in E. coli with other amino acids. ILV residues are predominantly found in hydrophobic pockets and cores, regions not well probed structurally by 1HN-1HN NOEs. For this reason, structural information derived from 1HN-1HN NOEs should be uniquely complemented by that derived from NOEs between ILV side chain protons. The well-characterized murine Grb2 N-SH3/SOSE proteinpeptide complex16 has been used to demonstrate the specificity of this labeling scheme. 1H/13C/15N-ILV residues have been incorporated into the perdeuterated and uniformly 15N-labeled N-terminal SH3 domain of mGrb217 (ILVN-SH3/SOSE). The labeling methodology consists of expressing the protein in minimal medium containing 95% D2O, [2H]-glucose, 15NH4SO4, and 1H/13C/15N-{Ile, Leu, Val} amino acids. The exact composition of the expression medium is provided in the supporting information. Glucose, instead of acetate,18 has been used as the sole carbon source to minimize potential scrambling of the labeled ILV residues with other residue types. With [1H]glucose, aromatic rings retain significant levels of protonation;19 [2H]-glucose is therefore required. The 2D constant-time 1H-13C HSQC on ILVN-SH3/SOSE (Figure 1) demonstrates the high selectivity in 1H/13C incorporation: no non-ILV residues are labeled. Furthermore, the spectrum contains no discernible deuterium isotope-shifted peaks associated with either methylene or methyl correlations. This indicates that most, if not all, of the ILV side chain groups are fully protonated. In line with previous observations,11 only 1520% fractional protonation of ILV CR carbons is observed in a 1D 13C-filtered proton spectrum (Figure 3 in supporting (13) Arrowsmith, C. H.; Pachter, R.; Altman, R. B.; Iyer, S. B.; Jardetzky, O. Biochemistry 1990, 29, 6332-6341. (14) Kay, L. E.; Bull, T. E.; Nicholson, L. K.; Griesinger, C.; Schwarbe, H.; Bax, A.; Torchia, D. A. J. Magn. Reson. 1992, 100, 538-558. (15) Bax, A.; Max, D.; Zax, D. J. Am. Chem. Soc. 1992, 114, 69236925. (16) Wittekind, M.; Mapelli, C.; Goldfarb, V.; Lee, V.; Meyers, C. A.; Mueller, L. Manuscript in preparation. (17) Lowenstein, E. J.; Daly, R. J.; Batzer, A. G.; Li, W.; Margolis, B.; Lammers, R.; Ullrich, A.; Skolnick, E. Y.; Schlessinger, J. Cell 1992, 70, 431-442. (18) Venters, R. A.; Calderone, T. L.; Spicer, L. D.; Fierke, C. A. Biochemistry 1991, 30, 4491-4494. (19) LeMaster, D. M. Prog. NMR Spectrosc. 1994, 25, 371-419.

+

+

Communications to the Editor

Figure 1. Expansion of the β/γ/δ region from the 2D gradientenhanced, constant-time 1H-13C HSQC spectrum on ILVN-SH3/SOSE. Only {Ile, Leu, Val} β/γ/δ correlations are observed in the 13C region upfield of 47 ppm; all such correlations are accounted for. Peaks labeled with an asterisk are unassigned but do not correspond to the known chemical shifts16 of any 1H/13C resonance pair in this protein. These extra peaks may therefore reflect some level of conformational heterogeneity within the protein-protein complex. Contours are plotted at a spacing of 1.6. The details pertaining to data collection are provided in the supporting information.

Figure 2. The superposition of the HCA II crystal structure (thin line) and the energy-minimized average HCA II structure (thick line) is presented in stereoview. The details of the simulated structure calculation for HCA II are described in the text.

information). The high level of deuteration at the ILV CR carbons should increase sensitivity for all amide-detected experiments, including the HC(CC)(CO)NH20 and the 13C/15Nseparated NOESY.21 The mechanism by which the R proton is replaced by a deuteron has been postulated to involve exchange with solvent deuterons (or protons).11 Such an exchange mechanism may allow expression of uniformly 15Nlabeled protein in which only the amide and R sites are substantially protonated. The 1H/13C resonances in large, type-specifically ILV protonated proteins (ILVproteins) can be assigned based on correlations observed in a 4D 13C/15N-separated NOESY (unpublished results). In this assignment process, the efficacy of the NOESY is enhanced by focused intraresidue spin-diffusion from the side chain protons out to their intraresidue and sequential amide protons. Although this assignment process is only minimally complicated by spectral overlap of Leu and Val methyl resonances, such overlap may prevent the unique assignment of NOEs between ILV methyl protons. To address this issue, we have also developed a protocol to express IVprotein, in which Leu residues are both 13C- and 2H-labeled. A constant-time 1H-13C HSQC spectrum on IVN-SH3/SOSE (Figure 4 in supporting information) confirms the applicability of this protocol in mitigating spectral overlap of Leu and Val methyl resonances. (20) Montelione, G. T.; Lyons, B. A.; Emerson, S. D.; Tashiro, M. J. Am. Chem. Soc. 1992, 114, 10974-10975. (21) Kay, L. E.; Clore, G. M.; Bax, A.; Gronenborn, A. M. Science 1990, 249, 411-414. Muhandiram, D. R.; Xu, G. Y.; Kay, L. E. J. Biomol. NMR 1993, 3, 463-470.

J. Am. Chem. Soc., Vol. 118, No. 28, 1996 6801 Analysis of a 3D 13C-edited/13C-separated NOESY (τm ) 150 ms, 1.4 mM) indicates an NOE cutoff distance of ∼5 Å between 13C-attached protons in ILVN-SH3/SOSE. Although the NOESY on ILVHCA II will experience decreased sensitivity due both to 13C labeling and to the concomitant use of 4D heteronuclearseparated spectroscopy, the increased dispersion is absolutely essential in order to accurately assign NOEs involving ILV methyl protons. In a 4D 13C/13C-separated NOESY on ILVHCA II, the absolute level of sensitivity is expected to approach that in the 4D 15N/15N-separated NOESY on 2H-HCA II.7 This expectation is based on a comparison of experimental 15N and 1H 2 N T2 relaxation times for H-HCA II (51 and 30 ms, respectively; τc ) 11.4 ns)22 with predicted 13Cmethyl and 1Hmethyl T2 relaxation times for ILVHCA II (46 and 48 ms, respectively).23 To demonstrate the benefit of this ILV labeling scheme to structure determination of large proteins, simulations were performed using theoretically observable aliphatic-aliphatic and aliphatic-amide NOEs in perdeuterated and uniformly 15Nlabeled ILVHCA II (29 kDa) derived from the crystal structure.24 Restraints were derived using NOE cutoff distances of 4.5, 4.0, and 3.5 Å for proton methyl-methyl, methyl-(non-methyl), and (non-methyl)-(non-methyl) interactions, respectively. In addition, cutoff distances of 4.5 and 4.0 Å were used for amidemethyl and amide-(non-methyl) interactions. To allow for inaccuracies in NOE intensities resulting from spin diffusion and differential relaxation, the upper bound of each distance restraint was set to 1.5 times the extracted distance. Intraresidue aliphatic-aliphatic NOE restraints and any knowledge of stereoassignments were excluded from the calculations. After combining these restraints (604 in total) with previously described 1HN-1HN NOE restraints7 (596 in total), an ensemble of 20 structures was generated by distance geometry and restrained dynamic simulated annealing.25 The N-terminus and several loops were ill-defined in these simulations, primarily because these regions contain no ILV and few side chain 1HN-containing residues, Excluding the N-terminal 16 residues in the HCA II crystal structure, the precision of the NMR ensemble is 3.1 Å, with the energyminimized, average structure superimposing on the target crystal structure with a backbone RMSD of 2.9 Å (Figure 2). If the disordered loops are excluded in the superposition, the RMSD decreases to 2.1 Å. This represents a dramatic improvement over the ensemble refined solely against 1HN-1HN restraints, whose RMSD from the crystal structure was 6.4 Å (for residues 21-260).7 The inclusion of torsion-angle and/or chemical-shift restraints may yield even further improvement. Acknowledgment. B.T.F. acknowledges Dr. Ron Venters of Duke University for useful discussions in the initial phases of this work. Supporting Information Available: Composition of ILVN-SH3 expression medium; acquisition details for the constant-time 1H-13C HSQC spectrum in Figure 1; 1D 13C-filtered proton spectrum indicating residual HR signal intensity in ILVN-SH3/SOSE; 2D constant-time 1H13C HSQC spectrum of IVN-SH3/SOSE (5 pages). See any current masthead page for ordering and Internet access instructions. JA9604875 (22) Venters, R. A.; Farmer, B. T., II; Fierke, C. A.; Spicer, L. D. Submitted for publication. (23) Model studies14 were used to predict T2(13Cmethyl) for ILVHCA II; experimental 1Hmethyl line widths for a 38.5 kDa ILV-labeled flavoprotein (τc ∼ 18-19 ns) (unpublished results) were used to predict T2(1Hmethyl) for ILVHCA II. (24) Liljas, A.; Kannan, K. K.; Bergsten, P.-C.; Waara, I.; Fridborg, K.; Strandberg, B.; Carlbom, U.; Jarup, L.; Lovgren, S.; Petef, M. Nat. New Biol. 1972, 235, 131-137. (25) Nilges, M.; Gronenborn, A. M.; Brunger, A. T.; Clore, G. M. Protein Eng. 1988, 2, 27-38.